Make.com Integration

CortexDB integrates with Make.com (formerly Integromat) to bring long-term memory to your visual automation workflows. Use the HTTP module to call CortexDB's REST API, or import the pre-built app module definitions.

Setup with HTTP Module

The fastest way to use CortexDB in Make.com is via the HTTP module.

1. Create a Connection

In your Make.com scenario:

1. Add an HTTP > Make a request module
2. Click Add next to the Connection field
3. Configure:
   • Connection name: CortexDB
   • Base URL: Your CortexDB server URL (e.g., https://api.cortexdb.ai)

2. Set Headers

For each HTTP module, add these headers:

| Header | Value | |---|---| | Content-Type | application/json | | Authorization | Bearer YOUR_API_KEY |

Operations

Remember (Store Content)

HTTP Module Configuration:

• URL: {{connection.base_url}}/v1/remember
• Method: POST
• Body type: Raw (JSON)

{
  "content": "The weekly sales meeting decided to increase Q2 targets by 15%.",
  "tenant_id": "sales-team",
  "scope": "meetings",
  "metadata": {
    "source": "make",
    "meeting_date": "2024-01-15"
  }
}

Recall (Retrieve Context)

HTTP Module Configuration:

• URL: {{connection.base_url}}/v1/recall
• Method: POST
• Body type: Raw (JSON)

{
  "query": "What were the recent sales decisions?",
  "tenant_id": "sales-team",
  "max_tokens": 4096,
  "min_confidence": 0.5
}

Forget (Remove Memories)

HTTP Module Configuration:

• URL: {{connection.base_url}}/v1/forget
• Method: POST
• Body type: Raw (JSON)

{
  "query": "user data for [email protected]",
  "reason": "GDPR deletion request",
  "tenant_id": "default"
}

Search (Find Episodes)

HTTP Module Configuration:

• URL: {{connection.base_url}}/v1/search
• Method: POST
• Body type: Raw (JSON)

{
  "query": "deployment events",
  "tenant_id": "ops-team",
  "source": "ci-cd",
  "namespace": "production",
  "limit": 20,
  "offset": 0
}

List Episodes (for Triggers)

HTTP Module Configuration:

• URL: {{connection.base_url}}/v1/episodes
• Method: GET
• Query parameters:
  • tenant_id: default
  • limit: 50
  • offset: 0

Example Scenarios

Google Sheets to CortexDB Memory

1. Watch Rows (Google Sheets): Trigger on new rows
2. HTTP Make a Request: POST to /v1/remember
   • Content: Row data formatted as text
   • Tenant: spreadsheet-data

AI Chat with Memory (using Webhooks)

1. Webhook: Receive incoming chat message
2. HTTP Make a Request: POST to /v1/recall
   • Query: {{webhook.message}}
3. HTTP Make a Request: POST to OpenAI API
   • Include recalled context in the prompt
4. HTTP Make a Request: POST to /v1/remember
   • Store the conversation exchange
5. Webhook Response: Return the AI response

Scheduled Memory Cleanup

1. Schedule: Run weekly
2. HTTP Make a Request: POST to /v1/forget
   • Query: Stale data criteria
   • Reason: Weekly cleanup

Slack Integration with Memory

1. Watch Messages (Slack): New messages in a channel
2. Router: Filter important messages (e.g., containing keywords)
3. HTTP Make a Request: POST to /v1/remember
   • Content: {{slack.text}}
   • Metadata: {"channel": "{{slack.channel}}", "user": "{{slack.user}}"}

Using the Custom App Definition

For advanced users, import the CortexDB app definition:

1. In Make.com, go to My Apps > Create a New App
2. Use the app definition from the cortexdb-make package
3. The app includes pre-configured modules for all CortexDB operations

REST API Reference

| Endpoint | Method | Description | |---|---|---| | /v1/remember | POST | Store content in memory | | /v1/recall | POST | Retrieve relevant context | | /v1/forget | POST | Remove memories | | /v1/search | POST | Search for episodes | | /v1/episodes | GET | List recent episodes | | /v1/health | GET | Health check |

Tips

• Use variables: Reference data from previous modules using Make.com's mapping panel
• Error handling: Add error handlers to your HTTP modules to gracefully handle CortexDB downtime
• Rate limiting: CortexDB handles concurrent requests well, but add a delay between bulk operations
• JSON parsing: After each HTTP module, use a JSON > Parse JSON module to extract response fields